The invention relates to a graph calculation optimization method based on heterogeneous FPGA data streams. An FPGA processing module is used for reading graph data Graph data stored in the shared storage module, and traversing the graph data based on the power law distribution characteristics of the graph data by virtue of the editability of the FPGA processing module to obtain at least one irregularity parameter of an irregular data stream existing in the graph data before the graph data is subjected to graph calculation Graph Procissing; and a scheduling module is used for distributing a part of the graph data matched with the CPU processing module to the CPU processing module for graph calculation according to a mode of matching the at least one irregular shape parameter with a preset access rule matched with the irregular shape parameter so as to obtain a CPU graph calculation result and distributing the rest of the graph data to the graph data with the parallel processing functionand the FPGA processing module performs graph calculation to obtain an FPGA graph calculation result. According to the method, the processing loads of the CPU processing module and the FPGA processing module can be balanced as much as possible in the graph data acceleration processing process.